Overclock.net › Forums › Industry News › Hardware News › [TT] MSI HD6970 Lightning overclocked review
New Posts  All Forums:Forum Nav:

[TT] MSI HD6970 Lightning overclocked review - Page 6

post #51 of 80
Quote:
Originally Posted by PoopaScoopa;13092937 
2nd - A real math lesson: 880/1015 = 13% Clock increase. 38/44 = 14% fps increase. Where did you learn to do math from?

Last thing about math. I think most people consider 2.2 to be 10% bigger than 2 and NOT 2/2.2 or 9.09% bigger like you seem to ...

Quote:
3rd - You didn't provide benchmarks of 929 to 1017. You provided a comparison from 880 to 1015. If you actually knew about OCing, you would know that bumping the clock from 1100Mhz down to 1000 wouldn't drop fps by an exact 9% as you're trying to claim it being completely linear all the way.
I already stated that I tested my DCII 570s beyond my current 1020Mhz using up to 1.212V and didn't see much of a performance gain and didn't feel comfortable with the high voltage or heat. There is a point at which further OCing will have severe diminishing returns as any good OCer would know. The Sapphire HD's fps won't scale up to 1100Mhz as well as the MSI Lightning for instance since it can handle more current.

OT, of course there is a wall for every GPU. But do you REALLY think 6970s hit a wall at 929MHz? This thread was about the review for the 6970 TFIII... I seriously doubt there is wall for those cards at 929MHz. I tried to show that by looking at a review of the card comparing stock performance (will be the same for all revisions of the card) to 1015Mhz (OC on the 6970 Lightning), but you dismissed that saying that the performance increase is not completely linear....

OF COURSE that is true. But your BC2 benchmark shows a ~1% increase for an increase of 929Mhz to 1017MHz. So are trying to say that in my AvP benchmark they had 15% FPS increase for 5.5% (880 to 929) increase in core clocks and 0% the rest of the way (929 to 1015). Really... that's what you think. It is not going to be exactly linear, but it will be close. You are not going to get serious diminishing returns with a ~6% increase core clocks on the 6970. If you want to claim that then show proof from another source. Your precious 570s on the other hand are clocked up approx. 30% over stock. So yeah, it makes a lot of sense that you are near the wall.

The OC scaling is depended on the hardware. My problem with you is that you are showing a driver problem in one game causing the card to scale poorly at high OCs and making general claims about the card. The same is not true in other games as I show. As for the other games you posted, I have no idea about the clocks.


tl;dr, Coming into a 6970 Lightning review thread, posting your pet benchmarks, and them trying to claim that the 6970 scales extremely poorly for clocks over 929MHz is really just silly.

Btw, your point about how 6970 CFX is much closer to 570 SLI than 580 SLI is true thumbsupsmiley.png
Edited by JCPUser - 4/12/11 at 3:28am
post #52 of 80
Quote:
Originally Posted by JCPUser View Post
tl;dr, Coming into a 6970 Lightning review thread, posting your pet benchmarks, and them trying to claim that the 6970 scales extremely poorly for clocks over 929MHz is really just silly.
Isn't that what fanboys try and do?.. He's coming to conclusions that 6970s can't scale well overclocked based on 1 benchmark. It's not even a fixed benchmark, so if the reviewer looked at more explosions in that run than he did on stock then obviously the average FPS would be lower, you get it?.. It's not fixed so it will never be exactly the same.
    
CPUMotherboardGraphicsRAM
i5-2500k @ 4300Mhz Biostar TP67XE Waiting to buy a new one 8GB 1600Mhz 7-8-7-21 1T 
Hard DriveCoolingOSMonitor
Samsung 840 SSD [250GB] Antec H20 620 Windows 7 x64 Acer S242HL 
PowerCase
Thermaltake 775W NZXT Phantom [White] 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
i5-2500k @ 4300Mhz Biostar TP67XE Waiting to buy a new one 8GB 1600Mhz 7-8-7-21 1T 
Hard DriveCoolingOSMonitor
Samsung 840 SSD [250GB] Antec H20 620 Windows 7 x64 Acer S242HL 
PowerCase
Thermaltake 775W NZXT Phantom [White] 
  hide details  
Reply
post #53 of 80
Quote:
Originally Posted by JCPUser View Post
Last thing about math. I think most people consider 2.2 to be 10% bigger than 2 and NOT 2/2.2 or 9.09% bigger like you seem to ...
I'm sorry, that was actually a 16% fps increase and 15% clock increase. I was tired and reversed the division.

Quote:
Originally Posted by JCPUser View Post
OT, of course there is a wall for every GPU. But do you REALLY think 6970s hit a wall at 929MHz? This thread was about the review for the 6970 TFIII... I seriously doubt there is wall for those cards at 929MHz. I tried to show that by looking at a review of the card comparing stock performance (will be the same for all revisions of the card) to 1015Mhz (OC on the 6970 Lightning), but you dismissed that saying that the performance increase is not completely linear....
There was a 6990 at 940Mhz posted alongside the Sapphire 6970 and it had a slight performance increase of the later. So 11Mhz did help it. But when 80Mhz was added to the Sapphire only 3 fps in 1920 and 1 fps in 2560 were added. I was pointing out that not all cards benefit past a certain point. If there was a benchmark comparing 1000Mhz to 1100 on the Lightning, it's highly unlikely that the gains will be as beneficial as 880-980 is.

Quote:
Originally Posted by JCPUser View Post
OF COURSE that is true. But your BC2 benchmark shows a ~1% increase for an increase of 929Mhz to 1017MHz. So are trying to say that in my AvP benchmark they had 15% FPS increase for 5.5% (880 to 929) increase in core clocks and 0% the rest of the way (929 to 1015). Really... that's what you think. It is not going to be exactly linear, but it will be close. You are not going to get serious diminishing returns with a ~6% increase core clocks on the 6970. If you want to claim that then show proof from another source. Your precious 570s on the other hand are clocked up approx. 30% over stock. So yeah, it makes a lot of sense that you are near the wall.
I didn't say 0% from 929 to 1017. I just said the gains wouldn't be as linear as going from 880-1015 are. A benchmark showing one card going from 929 to 1017 showed limited gains. There was a 3fps gain at 1920 and 1fps gain at 2560.

Quote:
Originally Posted by JCPUser View Post
The OC scaling is depended on the hardware. My problem with you is that you are showing a driver problem in one game causing the card to scale poorly at high OCs and making general claims about the card. The same is not true in other games as I show. As for the other games you posted, I have no idea about the clocks.
The 6000s only have problems at 5760x1080. It's not just in BC2 either. Every game except for a few benchmarks performed worse or equal to a 570 with 720 Megabytes of lesser frame buffer. Half the games did extremely poorly showing some serious scaling problems unlike their well performed results at 2560 and below.

Quote:
Originally Posted by JCPUser View Post
tl;dr, Coming into a 6970 Lightning review thread, posting your pet benchmarks, and them trying to claim that the 6970 scales extremely poorly for clocks over 929MHz is really just silly.

Btw, your point about how 6970 CFX is much closer to 570 SLI than 580 SLI is true
I didn't come into a 6970 Lightning thread claiming all 6970s scaled extremely poorly beyond 929Mhz. I came in saying they're more equal to a 570s than a 580s. I showed that it scales poorly at 5760x1080 in many games and does poorly as a single card but does extremely well in 2-way at 1920 and 2560 catching up to 570s. They may be weaker individually, but Nvidia's poor scaling and AMD's phenomenal scaling with the 6000 series makes up for it.

Like I said, in their price range, if you buy just one card, don't make it a 6970. If you buy 2, make it 6970s. If you're playing at 5760x1080 as I will be, make it 570s.

Quote:
Originally Posted by Razi3l View Post
Isn't that what fanboys try and do?.. He's coming to conclusions that 6970s can't scale well overclocked based on 1 benchmark. It's not even a fixed benchmark, so if the reviewer looked at more explosions in that run than he did on stock then obviously the average FPS would be lower, you get it?.. It's not fixed so it will never be exactly the same.
Going for whichever company makes the best card is apparently fanboyism. Even if that switches every year... The same scaling issues at 5760x1080 were reflected across all games even though they have an extra 720MB of frame buffer. The 6970 never caught up to a single 570 in any of the games except for a couple synthetic benchmarks.

The single player level used in testing is a cinematic sequence. Each run is identical. Otherwise the 5970 for instance wouldn't always end up with the same fps in each new review. Trying to dismiss a well established review site that has been reviewing components since 2001 just shows how limited your experiences into the community of OCing is.
Edited by PoopaScoopa - 4/12/11 at 9:42am
post #54 of 80
I just ordered the 6970 Lightning from the Egg, due to a favor being repaid from a friend my cost will be $95.86.

SCORE
My First Build
(15 items)
 
  
CPUMotherboardGraphicsRAM
I7-860 2.8/OC'ed 3.60 FPO/batch # L935B730 Asus P7P55d Evo MSI 580 TF II OC 8 Gig Corsiar DDR3 1600 XMP 
Hard DriveOptical DriveCoolingOS
WD Black 750 Samsung H50, pull/push Win7 Pro 64 
MonitorKeyboardPowerCase
HP 2509B 25" 1920x1080 Steelseries backlit Merc Corsiar TX650 Thermatake V9 Black 
MouseAudio
Rat 7 Asus Xonar DG 
  hide details  
Reply
My First Build
(15 items)
 
  
CPUMotherboardGraphicsRAM
I7-860 2.8/OC'ed 3.60 FPO/batch # L935B730 Asus P7P55d Evo MSI 580 TF II OC 8 Gig Corsiar DDR3 1600 XMP 
Hard DriveOptical DriveCoolingOS
WD Black 750 Samsung H50, pull/push Win7 Pro 64 
MonitorKeyboardPowerCase
HP 2509B 25" 1920x1080 Steelseries backlit Merc Corsiar TX650 Thermatake V9 Black 
MouseAudio
Rat 7 Asus Xonar DG 
  hide details  
Reply
post #55 of 80
Quote:
Originally Posted by Spct View Post
I just ordered the 6970 Lightning from the Egg, due to a favor being repaid from a friend my cost will be $95.86.

SCORE
Happy Overclocking. I ventured over from the green side and bought two of these very solid cards -- just couldn't help myself. Another review by G3D was posted in a different thread -- I have linked it below:

http://www.guru3d.com/article/msi-ra...ghting-review/

In that review they complain about throttling, so I decided some research on the subject. MSI made this card for people wanting to do LN2 suicide runs and break records so it didn't make sense for the card to be getting throttled.

One thing I found after research was a review done on xtreme systems:

http://www.xtremesystems.org/forums/...d.php?t=268825

They reveal the existence of an "OCP Unlocker" switch on the back of the card. This switch is not advertised by MSI nor is it completely clear what it changes about PowerTune (AMD's OCP). I believe that it will COMPLETELY disable PowerTune.

The other (safer) option for making sure the card doesn't throttle is the dual BIOS switch. Check out this MSI documents link below (PDF):

http://au.msi.com/download/Media/VGA...0Lightning.pdf

About half way down they mention that switching to the "performance" BIOS (BIOS 2) they 1) increase the CCC limits and 2) open the Powertune range by 2.5X times. I have not installed my cards yet (waiting for Bulldozer), but I think the range will go from 40 (-20% to +20%) to 100 (-20% to +80%). Furthermore, the 950MHz clock limit in CCC that is imposed for all 6970 by AMD should be raised as well. This feature seems to be only for 6970 TF III cards.


Summary: I recommend switching to BIOS 2 (Performance BIOS) if you are going to OC past 1000MHz so that you get a bigger powertune adjustment range thus keeping your card from throttling as that will kill any scaling at extreme clocks. This may have been happening in some of the benches presented in this thread with vanilla 6970 @ big OCs. If you are going to really go for it and put the card under water (or go LN2) then you may want to disable OCP completely on the back of the card.

Cheers.
Edited by JCPUser - 4/12/11 at 9:44pm
post #56 of 80
Quote:
Originally Posted by JCPUser View Post
Happy Overclocking. I ventured over from the green side and bought two of these very solid cards -- just couldn't help myself. Another review by G3D was posted in a different thread -- I have linked it below:

http://www.guru3d.com/article/msi-ra...ghting-review/

In that review they complain about throttling, so I decided some research on the subject. MSI made this card for people wanting to do LN2 suicide runs and break records so it didn't make sense for the card to be getting throttled.

One thing I found after research was a review done on xtreme systems:

http://www.xtremesystems.org/forums/...d.php?t=268825

They reveal the existence of an "OCP Unlocker" switch on the back of the card. This switch is not advertised by MSI nor is it completely clear what it changes about PowerTune (AMD's OCP). I believe that it will COMPLETELY disable PowerTune.

The other (safer) option for making sure the card doesn't throttle is the dual BIOS switch. Check out this MSI documents link below (PDF):

http://au.msi.com/download/Media/VGA...0Lightning.pdf

About half way down they mention that switching to the "performance" BIOS (BIOS 2) they 1) increase the CCC limits and 2) open the Powertune range by 2.5X times. I have not installed my cards yet (waiting for Bulldozer), but I think the range will go from 40 (-20% to +20%) to 100 (-20% to +80%). Furthermore, the 950MHz clock limit in CCC that is imposed for all 6970 by AMD should be raised as well. This feature seems to be only for 6970 TF III cards.


Summary: I recommend switching to BIOS 2 (Performance BIOS) if you are going to OC past 1000MHz so that you get a bigger powertune adjustment range thus keeping your card from throttling as that will kill any scaling at extreme clocks. This may have been happening in some of the benches presented in this thread with vanilla 6970 @ big OCs. If you are going to really go for it and put the card under water (or go LN2) then you may want to disable OCP completely on the back of the card.

Cheers.
Gotta love MSI's lightning models; can't beat 12 GPU VRMs.
post #57 of 80
Thats one of the factors in my purchase. The heat and power phases on this card are awesome. The headroom that this card should have is rediculous.

Looking forward to running benchies on stock then tweaking up to see whats what.

One other thing, when you see the GPU config,

From Guru3D:

'One other detail that you might find interesting is that when you look at the block diagram, you'll notice that the GPU pretty much looks like a dual-core processor. AMD calls this dual graphics engines. Anyway, have a peek at the block diagrams if at all interested."



http://www.guru3d.com/article/msi-ra...hting-review/2
Edited by Spct - 4/13/11 at 6:44pm
My First Build
(15 items)
 
  
CPUMotherboardGraphicsRAM
I7-860 2.8/OC'ed 3.60 FPO/batch # L935B730 Asus P7P55d Evo MSI 580 TF II OC 8 Gig Corsiar DDR3 1600 XMP 
Hard DriveOptical DriveCoolingOS
WD Black 750 Samsung H50, pull/push Win7 Pro 64 
MonitorKeyboardPowerCase
HP 2509B 25" 1920x1080 Steelseries backlit Merc Corsiar TX650 Thermatake V9 Black 
MouseAudio
Rat 7 Asus Xonar DG 
  hide details  
Reply
My First Build
(15 items)
 
  
CPUMotherboardGraphicsRAM
I7-860 2.8/OC'ed 3.60 FPO/batch # L935B730 Asus P7P55d Evo MSI 580 TF II OC 8 Gig Corsiar DDR3 1600 XMP 
Hard DriveOptical DriveCoolingOS
WD Black 750 Samsung H50, pull/push Win7 Pro 64 
MonitorKeyboardPowerCase
HP 2509B 25" 1920x1080 Steelseries backlit Merc Corsiar TX650 Thermatake V9 Black 
MouseAudio
Rat 7 Asus Xonar DG 
  hide details  
Reply
post #58 of 80
Quote:
Originally Posted by PoopaScoopa View Post
BC2 was just one of many games showed. Other games had similar performance problems at 5760x1080 as well and when they didn't, they did exactly equal or almost as good as the 570s. BC2 tends to be the standard just as BF3 will be in November. Some people may be more into RTS games and won't care though.

The original post was simply stating that 6970s are more equal to 570s than a 580 as someone else claimed in their post. As far as single card comparisons, the 560 did better than the 6970 in BC2, Metro 2033 and MW2. I showed benchmarks showing equal OC comparisons providing that information. I also said that the 6970 would be a better choice even though the 570 does just as good at 2560x1600 and below because it has more framer buffer to crank the AA up.

2nd - A real math lesson: 880/1015 = 13% Clock increase. 38/44 = 14% fps increase. Where did you learn to do math from?

3rd - You didn't provide benchmarks of 929 to 1017. You provided a comparison from 880 to 1015. If you actually knew about OCing, you would know that bumping the clock from 1100Mhz down to 1000 wouldn't drop fps by an exact 9% as you're trying to claim it being completely linear all the way.
I already stated that I tested my DCII 570s beyond my current 1020Mhz using up to 1.212V and didn't see much of a performance gain and didn't feel comfortable with the high voltage or heat. There is a point at which further OCing will have severe diminishing returns as any good OCer would know. The Sapphire HD's fps won't scale up to 1100Mhz as well as the MSI Lightning for instance since it can handle more current.

4th - There were other games besides just BC2 where the 6000s had serious performance issues at 5760x1080. When they didn't, they didn't do any better than the 570s which seems strange due to their extra 720MB of frame buffer.

I'm not claiming that people only play BC2 and that they only have problems with BC2; I was just pointing out that the 570s are more equal in performance/price than 580s are.
Not sure how you figured 570s do better in BC2

http://www.hardwareheaven.com/review...ty-vs-sli.html

3 6970s are more or less equal to 3 580s in BC2 at 5760x1080

2vs2 now



1vs1 they are pretty much tied.

and 1 more 3way for good measure


So they go from being tied at 1 card to getting murdered with 3 and the 570 is better how?
Server/Gaming Rig
(22 items)
 
  
CPUCPUMotherboardGraphics
Intel E5-2670 intel E5-2670 Asus Z9PE-D8 WS Visiontek R9 390 
RAMHard DriveHard DriveHard Drive
8x4GB DDR3 1333 ECC Registered Toshiba 5TB X300 Seagate 4TB HGST 4TB 
Hard DriveHard DriveHard DriveHard Drive
Sandisk 240GB Sandisk 480GB Samsung 830 128GB Toshiba 6TB X300 
Hard DriveHard DriveOptical DriveOS
Toshiba 5TB X300 Toshiba 5TB X300 WH16NS40 Windows 10 Pro 
MonitorMonitorKeyboardPower
Dell S1715H Dell u2311h blackwidow ultimate Kingwin Lazer platinum 1000w 
CaseMouse
Corsair 800D Deathadder Chroma 
  hide details  
Reply
Server/Gaming Rig
(22 items)
 
  
CPUCPUMotherboardGraphics
Intel E5-2670 intel E5-2670 Asus Z9PE-D8 WS Visiontek R9 390 
RAMHard DriveHard DriveHard Drive
8x4GB DDR3 1333 ECC Registered Toshiba 5TB X300 Seagate 4TB HGST 4TB 
Hard DriveHard DriveHard DriveHard Drive
Sandisk 240GB Sandisk 480GB Samsung 830 128GB Toshiba 6TB X300 
Hard DriveHard DriveOptical DriveOS
Toshiba 5TB X300 Toshiba 5TB X300 WH16NS40 Windows 10 Pro 
MonitorMonitorKeyboardPower
Dell S1715H Dell u2311h blackwidow ultimate Kingwin Lazer platinum 1000w 
CaseMouse
Corsair 800D Deathadder Chroma 
  hide details  
Reply
post #59 of 80
Oh god, not another stock comparison... Please learn what the O in OCN means.
post #60 of 80
Quote:
Originally Posted by PoopaScoopa View Post
Oh god, not another stock comparison... Please learn what the O in OCN means.
they are all stock vs stock its completely fair, there are no guarantees in overclocking.

please learn to use more than 1 source ffs
Server/Gaming Rig
(22 items)
 
  
CPUCPUMotherboardGraphics
Intel E5-2670 intel E5-2670 Asus Z9PE-D8 WS Visiontek R9 390 
RAMHard DriveHard DriveHard Drive
8x4GB DDR3 1333 ECC Registered Toshiba 5TB X300 Seagate 4TB HGST 4TB 
Hard DriveHard DriveHard DriveHard Drive
Sandisk 240GB Sandisk 480GB Samsung 830 128GB Toshiba 6TB X300 
Hard DriveHard DriveOptical DriveOS
Toshiba 5TB X300 Toshiba 5TB X300 WH16NS40 Windows 10 Pro 
MonitorMonitorKeyboardPower
Dell S1715H Dell u2311h blackwidow ultimate Kingwin Lazer platinum 1000w 
CaseMouse
Corsair 800D Deathadder Chroma 
  hide details  
Reply
Server/Gaming Rig
(22 items)
 
  
CPUCPUMotherboardGraphics
Intel E5-2670 intel E5-2670 Asus Z9PE-D8 WS Visiontek R9 390 
RAMHard DriveHard DriveHard Drive
8x4GB DDR3 1333 ECC Registered Toshiba 5TB X300 Seagate 4TB HGST 4TB 
Hard DriveHard DriveHard DriveHard Drive
Sandisk 240GB Sandisk 480GB Samsung 830 128GB Toshiba 6TB X300 
Hard DriveHard DriveOptical DriveOS
Toshiba 5TB X300 Toshiba 5TB X300 WH16NS40 Windows 10 Pro 
MonitorMonitorKeyboardPower
Dell S1715H Dell u2311h blackwidow ultimate Kingwin Lazer platinum 1000w 
CaseMouse
Corsair 800D Deathadder Chroma 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [TT] MSI HD6970 Lightning overclocked review